RE: a need for speed

["Lim Hock-Chai" <Lim.Hock-Chai@xxxxxxxxxxxxxxx>]

Actually, we already have a process for this.  It takes 3 hours+ to run.

I plan to rewrite to increase the speed.  Hence the question.

There is actually another twist to this process.  See below:

Our telephone master file is splitted across two AS400.  One has 22M+ and the 
other has 18M+ records.  The summary file needs to combine the record from the 
two AS400.  To copy such a big file across to another system, index it and run 
the process is probably too time consuming.  The plan is to create summary file 
on each system first, then somehow combine them.  Not sure if I can beat the 3 
hours time of the existing process.  Is going to be fun and probably will need 
some helps from you folks again.

thanks.



-----Original Message-----
From: rpg400-l-bounces@xxxxxxxxxxxx
[mailto:rpg400-l-bounces@xxxxxxxxxxxx]On Behalf Of Joel Fritz
Sent: Thursday, August 18, 2005 7:20 PM
To: RPG programming on the AS400 / iSeries
Subject: RE: a need for speed



One more thing to consider.  If this is a one off report, the easiest to
code approach is probably the best unless performance is abysmal.  If
it's a regular production job it kind of depends on how long both take.
If the run time is in hours it's probably worth optimizing it.  If it's
in minutes both ways it depends on how frequently it's run and whether
users are waiting with bated breath.

I wish I could remember the break even point between random and
sequential I/O, but I know it's completely counter intuitive. 

<><><>-----Original Message-----
<><><>From: rpg400-l-bounces@xxxxxxxxxxxx
<><><>[mailto:rpg400-l-bounces@xxxxxxxxxxxx] On Behalf Of Tom Liotta
<><><>Sent: Thursday, August 18, 2005 5:04 PM
<><><>To: rpg400-l@xxxxxxxxxxxx
<><><>Subject: RE: a need for speed
<><><>
<><><>AFAIK, Booth's method would "work", but Scott's _could_
<><><>easily be faster. The ordering would be the same for both
<><><>and grouping would be defined the same. Scott
<><><>positions-to and READs the first and last record in every
<><><>group giving lowest and highest suffix (as long as
<><><>ordering and grouping are correct) in the group. Booth's
<><><>method would give the lowest suffix at L1-Detail time and
<><><>the highest at L1-Total time.
<><><>
<><><>However, Scott has a point when the average number of
<><><>records in groups exceeds some number. If records are
<><><>1000 bytes in size and groups average say 50000 records,
<><><>then maybe it's not such a good idea to read individual
<><><>records. But if groups average 5 records, then it hardly
<><><>matters whether records are all read or not -- most will
<><><>be physically read anyway and there'll be no READPs nor
<><><>any other complicating code paths internally.
<><><>
<><><>Tom Liotta
<><><>
<><><>rpg400-l-request@xxxxxxxxxxxx wrote:

*****************************
NOTICE:
All e-mail sent to or from this e-mail address will be received or otherwise 
recorded by
The Sharper Image corporate e-mail system and is subject to archival, 
monitoring,
and review by and/or disclosure to Sharper Image security and other management.
This message is intended only for the use of the addressee and may contain 
information
that is privileged and confidential.

The contents of this message may contain personal views which are not the views 
of The Sharper Image.
If you are not the intended recipient, dissemination of this communication is 
prohibited.
*****************************